Auto-approve settings

Define the boundaries of what Test Companion is allowed to do without asking for your permission first.

Auto-approve settings is where you define the boundaries of what Test Companion is allowed to do without asking for your permission first. By default, the AI pauses and requests approval before performing most actions, reading files, editing code, running terminal commands. Auto-approve lets you selectively remove these approval gates for actions you are comfortable with, so the AI can work faster.

How to open Settings

1. Click the gear icon (⚙) in the top-right corner of the Test Companion panel.

   

2. Select the Auto-Approve Settings tab.

   

3. Make the necessary changes and click Save to apply them.

Enable Auto-Approve

This is the overall on/off switch. If you uncheck this, none of the individual action approvals below take effect — every action requires manual approval. This is useful when you want to temporarily lock everything down (for example, before a demo or when working on a production configuration).

Actions

Following the master toggle, you will find a list of individual actions that you can auto-approve. Each one controls a specific category of action that the AI may need to perform during a task.

A button labelled Enable all appears in the top-right corner of this section. Clicking it checks every action at once which is useful during initial setup if you want maximum autonomy and can then selectively disable individual actions.

Read project files

Reading project files is typically a low-risk action. The AI needs to read your code to understand your project structure, analyze existing tests, and generate new code that is consistent with your codebase.

Example: When you ask “Generate test cases for the checkout page,” the AI reads your existing page objects, utility files, and configuration to understand the patterns your project uses before writing new test code.

Read all files (sub-option)

This is a broader permission. When enabled, the AI can read files outside your current project folder — for example, a global configuration file in your home directory, or a shared library installed elsewhere on your machine.

When to enable: If you have shared test utilities, environment configuration files, or documentation located outside your workspace that the AI needs to reference.

When to keep disabled: In most cases. If the AI only needs to work within your project, the default workspace-scoped permission is sufficient and more secure.

Edit project files

This is where the AI’s output becomes tangible. It creates new test files, updates existing code, and modifies configuration. If you are using default interaction mode and want the AI to work independently, this should be enabled.

If you prefer to review every change before it is written, disable this. The AI will show you the proposed changes and wait for your approval.

Example: The AI detects that a test is failing because of a renamed CSS selector. With “Edit project files” auto-approved, it updates the selector in your page object file immediately. Without auto-approve, it shows you the proposed change and waits for you to click “Approve.”

Edit all files (sub-option)

Execute safe commands

Test Companion uses a built-in safety classifier to evaluate terminal commands before execution. Commands categorized as “safe” include read-only or easily reversible operations, such as:

• npm test or pytest (running tests)
• git status or git log (checking repository state)
• ls, cat, grep (viewing files and output)
• npm install (installing dependencies)

Even with this setting enabled, Test Companion always requires your approval for commands the classifier considers potentially destructive. These include:

• rm -rf or any recursive deletion commands
• git push or git reset --hard (irreversible Git operations)
• Commands that modify system-level configuration
• Commands containing pipes to potentially dangerous utilities

Example: You ask the AI to “Run the test suite and show me which tests fail.” With “Execute safe commands” enabled, the AI runs pytest --tb=short directly, displays the output, and summarizes the failures. Without this setting, it would show you the command first and wait for approval before running it.

When to enable this

• During active development when you are frequently running tests and builds
• When you trust the AI’s safety classifier and want faster iteration cycles
• When using workflow templates that involve predictable, safe terminal operations

When to keep disabled

Keep this disabled in the following situations:

• When working with production databases, deployments, or other sensitive infrastructure
• When you are new to Test Companion and want to understand what commands the AI runs
• If your organization’s security policy requires manual approval of all terminal executions

Recommended configurations by role